As one of the main drivers of Python at Netflix, this one is near and dear to my heart. Let me know if you have any questions, or come visit our booth!
Posts like these are great on a tech level, but they're also good recruitment - Seeing Python, along with hackers that I respect (like you) at Netflix make me at least the littlest bit tempted to go there ;)
But in all seriousness, we like to share what we do because sharing is cool and it attracts other people who like to share cool technology, which are the kind of people we like to work with.
P.S. I really enjoyed your talk at QCon SF last year!
Ruby and Python are similar in terms of execution speed and developer productivity. If your work requires something that Python has and Ruby doesn't(say numpy), then it might make sense; otherwise what would you accomplish by switching from Ruby to Python?
Simpler syntax?
Wakari (http://wakari.io) is nearing 1.0 and we're about to turn on paid monthly plans for beefier nodes and GPU nodes, and we have also been in talks with several companies about in-house private deployments.
Give us a shout or shoot me an email at pwang (at) continuum.io! :-)
There was many interesting things listed at the blog post. I have been planning a system similar in idea to Sting, your OLAP server (?).
I'm planning a light RPC style system on top of Pandas dataframe objects, transferring HTTP requests to Pandas method calls. Memory usage of such a system with our data is too expensive if all data is in use. So, I'm investigating possible caching and database solutions.
Could you perhaps elaborate on the implementation of Sting's in memory database?
[posted from yet another HN account]
So now that's +2. :)
The idea of marrying "slow" Python and OLAP is disturbing my sleep!
Hit me up on email (it's in my profile).
b) python does have first class functions?
c) threading still happens, you just can't take advantage of multiple cores. This isn't as big of a problem as you'd think though, you can still write very fast code.
Also, packaging is a whole different can of worms (but hopefully will be sorted by distutils2!).
Python lets you pass functions other ways than using lambdas, you know. If you've defined a function with the standard "def foo" style, you can pass "foo" as an argument.
Even the standard "1-line lambdas LOL" limitation can be circumvented by defining the function elsewhere and making the lambda a call to it.
I have been building large projects in Python for the last 10 yeas and never really thought "Hmm I wish I had more powerful lambda functions". You can always just have regular function and pass that around. Why do you need lambdas so much?
First class functions simply mean that a language treats functions as it would nearly any other data type, namely that they may be passed as parameters, bound to variables, and returned from other functions.
Python's poor handling of lambda expressions do not invalidate that functions are first-class citizens.
Java shines at the huge end, but for most everything else python is a more productive choice as it is so much more concise. Perhaps you are not yet using it optimally?